International Business Machines Corporation

Eugene Johnson

Robert Bradley Bennett

Edward Joseph Bendert

Robert Michael Nugent

Frank Bernhard Schmuck

James M Showalter

Susan Marie Farrell

Thomas Eugene Engelsiepen

Mary E Vendryes

Radha R Kandadai

Terrence Lee Woodnorth

Frank J Mangione

Robert Micheal Nugent

Eve Suzanne Berman

Brian Paul Dixon

In a multinode data processing system capable of partitioning operations a method is provided for adding nodes to a configuration of nodes in a manner which node quorums are employed but in which two quorum values (ideal and enforced) are employed and in which old nodes using a given file system are distinguished from new nodes. The method prevents even the transient use of inappropriate quorum values.

Dynamic quorum adjustment

Selected server operations that affect objects in a distributed computing system can be off-loaded from servers at which the objects are stored to other servers without the requirement of vertical partitioning of the affected objects and without off-loading entire affected objects. A client environment process that requests an operation on an object is notified of a task server to which selected off-load operations should be sent. The client preferably stores the task server identifier and thereafter sends such operation request directly to the identified task server. The object metadata information can be stored in the client environment, if desired. The object metadata at the owning repository server is maintained, if affected by the requested operation. A single task server can perform off-loaded functions from several other repository servers at the same node and at other nodes, and in that way reduce the workload of many servers. The functions that can be off-loaded include named pipe functions and byte range file lock operations.

Operation-partitioned off-loading of operations in a distributed environment

A client-server distributed computer processing system provides staged byte range locking with three levels of lock complexity and contention. When locking activity for a data object involves a plurality of clients, locking is managed centrally in the server operating environment. When only a single client is involved in requesting a lock on a data object such as a data file or portion of a data file, the requesting client can be permitted to manage the locking locally in the requesting client operating environment. Client-managed locking involves two stages of locking complexity on an object-by-object basis. The first stage of client locking involves only a log of current lock activity and handles relatively simple locking. When locking becomes more complex, the client locking is migrated to a second local locking stage, where more complex data structures can support lock fragmentations that affect various portions of the data object, such as partially overlapped byte ranges. Migration between the three stages of locking uses a high level interface rather than migrating implementation-dependant structures and thereby avoids implementation dependencies and local optimization restrictions. The three-stage locking scheme optimizes performance for the more common cases of lock management, where there is minimal contention and less complexity or overlap in locking scope.

Byte range locking in a distributed environment

A computer system having a shared disk file system running on on multiple computers each having their own instance of an operating system and being coupled for parallel data sharing access to files residing on network attached shared disks. Method are provided for use as a parallel file system in a shared disk environment by use of scalable directory service for the system with a stable cursor, a segmented allocation map. Dynamic prefetch and cached balance pools for multiple accesses improve the system. Extended file attributes are used for implementation of Access Control Lists in a parallel file system. Improvements to caching and cache performance developments balance pools for multiple accesses. A metadata node manages file metadata for parallel read and write actions. For our system, tokens are used for metadata node selection and identification, and we have enhanced token modes for controlling file size, as well as smart caching of byte range tokens using file access patterns and a byte range lock algorithm using a byte range token interface. Locking techniques reduce the overhead of a token manager which is also used in the file system recovery if a computer participating in the management of shared disks becomes unavailable or failed. Synchronous and asynchronous takeover of a metadata node occurs for correction of metadata which was under modification and a new computer node to be metadata node for that file. Locks are not constantly required to allocate new blocks on behalf of a user. We have provided recoverable local shares for Quote Management. A quota server and a quote client are provided for each file system, and a quotacheck utility allows parallel environments with minimal interference to other applications which require the use of quotas.

Parallel file system with extended file attributes

A client/server computer system manages shared files. A client includes a data cache and an associated cache manager, and executes a client application that requests data from a shared file. In response to the request, the client cache manager determines if the data is currently stored in the data cache, and if so, grants the client access to the data in the cache. However, if the data is not currently stored in the cache, the client cache manager sends to the server a request for multiple consecutive blocks of data beginning with the first block containing the data requested by the client from the shared file. The server includes a token manager which receives the request, and in response (a) awards the token for a first data block specified in the request regardless of contention for the first data block and (b) awards tokens for all other consecutive blocks following the first block of the multiple blocks held by a client who also holds the token for the first block. However, if another client currently holds a contending token for one of the consecutive blocks other than the first block but does not also hold the token for the first block, then the token management process awards only the tokens for the first block and all blocks between the first block and the one block.

System for awarding token to client for accessing first data block

A system and method efficiently manage clones of an object group. A base storage area contains metadata for each of multiple objects in the object group. The metadata for each object identifies data within each object. The system receives a request to clone a base storage area, and in response, the system creates a clone storage area containing an identification of the base storage area but not the metadata. If there is a subsequent request to update a portion of one of the objects, then the system copies the portion, copies the metadata into the clone storage area for the one object, makes the update to the object copy and changes the metadata in the base storage area to identify the portion copy instead of the original portion. If there is no subsequent request to update any of the multiplicity of objects, then the system does not copy the metadata for any of the multiplicity of objects into the clone storage area.

Creation of clone storage area with identification of base storage area

A computer system manages first and second different file systems. The system comprises first and second, different file managers having first and second different APIs for the first and second file systems, respectively. The system also comprises a hybrid file manager for managing the second file system based on requests made using the first API. The hybrid manager applies rules from both APIs to control atomicity of requests, access concurrency, and data consistency rules. The hybrid rules prevent violation of the rules of either API.

Growing community of inventors

Austin, TX, United States of America

Eugene Johnson